10/10 would do it again, except this time we may pay SiFive or someone like that for something requiring less "customization".
But, I should also mention, we only use the RISC-V cores as a pre/post-processor, scheduling engine, service processor. We have custom hardware that does the bulk of the inference math (we are a convolutional accelerator with a number of constraints traded off for speed and power). The fabric itself is driven by engines that are programmed by the scheduling engine (RISC-V).
Happy to answer more specific DMs.